linux-kernel - Re: [PATCH v3 3/5] mm, thp: introduce FOLL_SPLIT

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20190613141615.yvmckzi3fac4qjag@box>
Date:   Thu, 13 Jun 2019 17:16:15 +0300
From:   "Kirill A. Shutemov" <kirill@...temov.name>
To:     Song Liu <songliubraving@...com>
Cc:     LKML <linux-kernel@...r.kernel.org>,
        "linux-mm@...ck.org" <linux-mm@...ck.org>,
        "namit@...are.com" <namit@...are.com>,
        "peterz@...radead.org" <peterz@...radead.org>,
        "oleg@...hat.com" <oleg@...hat.com>,
        "rostedt@...dmis.org" <rostedt@...dmis.org>,
        "mhiramat@...nel.org" <mhiramat@...nel.org>,
        "matthew.wilcox@...cle.com" <matthew.wilcox@...cle.com>,
        "kirill.shutemov@...ux.intel.com" <kirill.shutemov@...ux.intel.com>,
        Kernel Team <Kernel-team@...com>
Subject: Re: [PATCH v3 3/5] mm, thp: introduce FOLL_SPLIT_PMD

On Thu, Jun 13, 2019 at 01:57:30PM +0000, Song Liu wrote:
> > And I'm not convinced that it belongs here at all. User requested PMD
> > split and it is done after split_huge_pmd(). The rest can be handled by
> > the caller as needed.
> 
> I put this part here because split_huge_pmd() for file-backed THP is
> not really done after split_huge_pmd(). And I would like it done before
> calling follow_page_pte() below. Maybe we can still do them here, just 
> for file-backed THPs?
> 
> If we would move it, shall we move to callers of follow_page_mask()? 
> In that case, we will probably end up with similar code in two places:
> __get_user_pages() and follow_page(). 
> 
> Did I get this right?

Would it be enough to replace pte_offset_map_lock() in follow_page_pte()
with pte_alloc_map_lock()?

This will leave bunch not populated PTE entries, but it is fine: they will
be populated on the next access to them.

-- 
 Kirill A. Shutemov